Modelling Microwave Backscattering from Sea Ice for Synthetic-Aperture Radar Applications
The spaceborne synthetic-aperture radar (SAR) is considered one of the key instruments for monitoring the ice cover in polar oceans and regional seas. This thesis is concerned with applications of electromagnetic scattering theory and SAR system theory for modelling the response of different types of sea ice in SAR imagery. The modelling is an important tool for improving the interpretation of SAR images. The emphasis has been on the frequency (5.3 GHz), polarization (vertical), and incidence angles (20 ° -26 °) used by the SAR onboard the first European Remote Sensing Satellite (ERS-1), which has been in operation since 1991.
A SAR image simulator has been developed, which is able to closely simulate the first and second order statistics of the SAR image over a natural distributed target. It has been used to evaluate sea ice classification algorithms when applied to SAR images with different spatial and radiometric resolution.
A number of field experiments in the Arctic Ocean and the Baltic Sea have been carried out involving coincident radar backscattering and surface characterization measurements with the objective to improve our understanding of the scattering physics involved. Surface scattering is modelled using accurate measurements of small-scale surface roughness with a laser profiler, which are input to the integral equation model (IEM). Volume scattering from air inclusions in the ice is estimated by a Rayleigh scattering model, which neglects multiple-scattering and dense medium effects.
A new scattering model is formulated for deformed sea ice which consists of randomly oriented ice blocks. It is shown that the mean backscattering is independent of the block size distribution provided that sizes and slopes are independent. The model includes scattering from both the upper and lower sides of the blocks and it is shown that a superimposed surface roughness is of little importance. The model has been compared with numerical scattering computations using the method of moments. Good agreement is obtained, except when the dielectric loss factor is low, in which case the block model underestimates the scattering by neglecting multiple scattering within the blocks.
Backscattering from multi-year ice in the Arctic Ocean has been measured using a 5.4 GHz radar scatterometer. The backscattering is found to increase during the fall freeze-up period, which is explained by the scattering model to be caused by an increased scattering from a porous upper layer as the liquid water freezes. The results show that the ERS-1 SAR is sensitive to upper layer ice temperatures close to the melting point.
Backscattering from thin Arctic sea ice has been measured using the ERS-1 SAR and is found to be highly variable. Due to variations of surface roughness and dielectric constant, a variation of about 10 dB is observed, which is successfully modelled with the IEM surface scattering model. Variations in surface roughness is caused by the presence of frost flowers, whereas variations in surface salinity and temperature give large dielectric fluctuations. Thin ice may therefore be confused with both thicker first-year ice and multi-year ice in ERS-1 SAR images.
Experiments in the Baltic Sea during three consecutive winters have been performed with measurements of both level and deformed sea ice. The backscattering in ERS-1 SAR images is found to be determined by the millimetre to decimetre scale surface roughness of the ice surface, except in regions with very low salinity. The dominance of surface scattering is used to formulate an inverse model, which estimates the ice surface roughness from ERS-1 SAR images during dry snow conditions.